i think the real issues about user interface decisions come down to making the interface useful. in an ideal UI, as in Bryce, nearly every feature of an application is accessible, obvious, and unconfusing. That is to say, every single element in a UI does what it looks like it's supposed to do, and the user finds the element to do a given task easily.
Part of that is a predictable UI. Thats' why apple doesn't want you to mess with the Aqua. Users get used to the way applications look and react.
Getting all of the features in an application in reach of the user is alot harder though. I'm presently neck deep in that issue, designing an Appointment application. There are a great many ways of displaying and collecting information, but it won't do to show all of that information at one time. My application needs to show only fragments of the information, but also make it possible to switch from fragment to fragment, and do so intuitively.
I find myself asking a fairly small number 1. Can i design this element more simply? 2. is it easy to locate and use this element? 3. does this element do everything that it needs to?
i think these questions are pivotal to a good UI. it makes it easy to learn to use the application, and it makes it easy for savvy users to use the application to it's greatest extent.
A good example of this kind of design is, ironically, Interface Builder. All of the tools in IB are grouped logically in two palets, and their meaning is immediately obvious. IB is only usefull to users that know what to do with nib files, but if you are savvy, then the only obstruction to using IB is your own creativity in Interface Design. The Tool itself is nearly transparent to its use